In this paper an efficient automatic method for robust segmentation of finger vessel-network and vein pattern extraction from infrared images acquired by a low-cost monochrome or multichannel camera, is proposed. After brightness normalization, the fingerprint lines are eliminated using the 2D dimensional discrete wavelet transformation. A set of twelve directional kernels is constructed, based on a dyadic wavelet transform, for each scale and is used to enhance the directional properties of veins. From maximum filters’ response along scale and direction, a neighborhood thresholding derives a binary segmented image to produce reliable patterns of finger veins. A post-processing module is used in case where low-quality images are to be segmented. Preliminary evaluation experiments of the proposed method demonstrate a number of advantages, compared to recently published methods.